Long-term, stable behavior of local field potentials during brain machine interface use.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Local field potentials (LFPs) and multi-unit spikes (MSPs) are candidate signals for brain-machine interfaces (BMIs).
• Previous research suggests LFPs may offer robust and long-lasting control signals.
• The long-term stability of these neural signals for BMI control requires further investigation.

Purpose of the Study:

• To assess the long-term stability of LFPs and MSPs as control signals for biomimetic brain-machine interfaces.
• To evaluate the performance and stability of LFP-based and MSP-based BMIs in non-human primates over extended periods.
• To determine if neural representations for BMI control remain consistent over time.

Main Methods:

• Two monkeys were trained to control a computer cursor using both LFP- and MSP-based biomimetic BMIs.
• Performance accuracy and stability were monitored over 11 months (LFP-BMI) and 6 months (MSP-BMI).
• Linear decoders were trained in each session to predict cursor velocity from single LFP features or MSPs, and then tested on the final session data to assess signal stability.

Main Results:

• Both LFP- and MSP-based BMIs enabled highly accurate cursor control.
• This high performance was sustained for 11 and 6 months, respectively, without requiring adaptation or retraining.
• A significant number of individual LFP features and MSPs demonstrated stable, high correlations with cursor velocity throughout the study duration.

Conclusions:

• LFPs and MSPs are stable neural signal sources for long-term brain-machine interface control.
• The motor cortical representations used for BMI control can remain stable over many months.
• These findings support the potential of LFPs and MSPs for developing robust and enduring BMI applications.